BentoML

After `bentoml build` step, you should be able to find the Dockerfile generated at
```cat "$(bentoml get &lt;bento&gt;:&lt;tag&gt; -o path)/env/docker/Dockerfile"```
as well as all the artifacts, that should give you enough control to use kaniko to build the dockerfile even without using `bentoml containerize`

you can refer to this article which is written by one of our user
<https://valohai.com/blog/building-a-solution-based-on-machine-learning/>

I think we just need to support kaniko as one of the OCI builder backend, but i don’t really see the point of having this since `yatai-image-builder` can already handle building within k8s cluster

imho kaniko shouldn’t be used anywhere else except k8s environment :slightly_smiling_face:

If you are just running kaniko within a k8s cluster, then I think you can write a custom OCI builder if you want to stay in Python, but <@U03UY1JB269> solution should suffice